Legal document summarization improves with lattice-based consolidation
LexLattice: Multilingual Extractive Summarization via Neural Cellular Automata on Document Hierarchies
Computation and LanguageMachine LearningNeural and Evolutionary Computing
Summary
Summarizing legal texts accurately is challenging because important information is spread across different parts of the document. The authors developed LexLattice, a tool that organizes the document's structure into a two-dimensional grid and uses a type of neural network called cellular automata to combine information before choosing what to include in the summary. This method works well across many languages, even those it hasn't seen before, because it focuses on the meaning rather than the exact words. LexLattice achieves better summary quality than bigger models that rely on extensive language-specific tuning.
What this means in practice
- •For legal technology developers: Integrate multilingual legal summarization that traces summaries back to original document sections for trustworthy outputs.$Commercial implications: Enables commercial legal AI products offering reliable, multilingual summaries directly linked to source texts for law firms and legal service providers.
- •For government document managers: Use hierarchical consolidation methods to create consistent summaries from complex multilingual legal documents with reduced computational overhead.
Authors
Sujay Uday Rittikar, Sheela Ramanna
Abstract
Faithfulness is a central concern in legal text summarization, which motivates extractive approaches that select verbatim content traceable to its source. Such methods typically rank paragraphs or other structural units in isolation, yet give little attention to consolidating evidence that is distributed across, and shares salience between, distant parts of a document. We introduce LexLattice, an extractive summarizer that reifies a legal act's hierarchy as a two-dimensional semantic lattice and consolidates over it with a masked 2D neural cellular automata before selection. LexLattice attains state-of-the-art ROUGE across all 24 languages of EUR-Lex-Sum in both multilingual and cross-lingual settings, surpassing instruction-tuned baselines with billions of parameters, despite concentrating all trainable capacity in a 1.8M parameter consolidator over a frozen multilingual encoder. A consolidator trained only on high-resource languages further transfers to unseen languages with near-lossless retention (0.99), indicating that the model operates on language-agnostic semantic geometry rather than surface form. Our results position explicit consolidation over document structure as a compact and traceable alternative to scale for multilingual legal summarization.